Repository navigation
viewing an array with a fieldless structured dtype does not update the dtype: #14344
Description
As discovered in #14310 , viewing an array with a fieldless structured dtype does not update the dtype/itemsize:
>>> a = np.zeros(2, dtype={'names':[], 'formats':[], 'offsets':[], 'itemsize':8})
>>> b = a.view(np.dtype([]))
>>> b
array([], shape=(2, 0),
dtype={'names':[], 'formats':[], 'offsets':[], 'itemsize':8})
>>> b.dtype == np.dtype([])
False
Activity
eric-wieser commentedon Aug 23, 2019
In my opinion, the expected behavior here is to throw an exception, since the new array would need to be 8/0 times longer.
ahaldane commentedon Aug 23, 2019
Yeah, I was just thinking about that too.
Alternatively, we could add a special case for size-0 arrays, which can always be viewed as any dtype. But I'm not sure I like special cases.
Edit: That is in reference to the case in #14310 where I wanted to view a size-0 array as a fieldless dtype.